Methodology

How a shield gets earned.

We say our rating criteria are published, dated, and versioned before we apply them. This page is that publication. If a standing on the map ever looks arbitrary, this is the page that should settle it, or the page that should change.

There is a name in the research for the thing this page exists to catch: ethics washing, publishing a set of high-minded principles for reputational credit while doing nothing that would make them cost anything to break. It works precisely because principles are cheap and enforcement is not. Ethics asks what is right; governance builds the structures, roles, and evidence trails that make an answer to that question actually operate; compliance is the legal floor underneath both. A shield on this map is our attempt to check which of those three a given credential is actually offering, whatever its landing page implies.

Three different jobs, on purpose

Values, operational practice, and enforceable law are not the same question, and this site keeps them on different pages because collapsing them into one is how ethics washing happens. The standard page states the values: the same ground UNESCO’s Recommendation on the Ethics of AI names as human dignity, privacy, transparency, fairness, accountability, and human oversight.

• The gap report and the practice shelf are the operational layer: the day-to-day judgement calls, templates, and evidence that turn a stated value into something a person actually did.
• The statute is the enforceable layer: the regulation that can fine an organisation regardless of what it claims to believe.
• A shield on the map tells you where a credential sits in the middle layer, and it was never a substitute for the other two.

The three shields, exactly

Every entry on the map gets one of three standings. Nothing else moves it, not the size of the provider, not how long they have existed, not how they describe themselves.

How we score a course

A course never gets a shield, but it does get a scorecard in its detail drawer, three plain marks rather than one composite number, since a single score would flatten distinctions worth keeping separate: how much the course actually covers, whether finishing it requires real assessed work or just clicking through, and what you walk away with, a certificate that carries some weight or one that carries none. Each mark is light, moderate, or rigorous, and none of them is a stand-in for accreditation, which only the three shields above can carry.

Every scorecard also says plainly whether it is based on our own team actually completing the course, or on the provider’s published syllabus, materials, and public reviews without full completion. We would rather sit through more of them ourselves than pretend a syllabus read like a finished course, but time is finite and the map is long, so where we have not completed something, we say so on the entry itself rather than let a confident-sounding score imply we did.

Anyone who has actually taken a course can tell us if a mark is wrong, from the scorecard itself. That feedback never moves a score by itself: a person reads it, checks it against the course, and only then decides whether anything changes, dated as everything else here is. A single submission, or a cluster of similar ones arriving together, gets treated as a reason to look closer, not as a vote, which is what keeps a provider polishing its own listing, or a rival trying to sink someone else’s, from working.

What we check, every time

Before an entry gets a standing, or keeps one, six questions get asked against primary sources, not the provider’s own claims:

• Is there a named accreditation or standards body behind it, and can we find that body’s own public register listing this scheme?
• Can someone actually sit or earn it today, right now, not “soon” or “in pilot”?
• Who governs it, and is that governance disclosed anywhere we could find without asking?
• What does it cost, and is that cost public, or did we have to ask a salesperson?
• Where it claims alignment with EN 18274 or a similar competence framework, does the syllabus actually cover those domains, or just use the language?
• Is it free and academic, or commercial? The two are grouped and labelled separately on the map, never blended.

Dated and versioned, in practice

Every entry carries the date it was last checked against its primary source, shown on the map and in each detail drawer. “Versioned” means a standing is never silently edited: when a status changes, the old status, the new one, the date, and the reason are recorded, not overwritten. There is no draft state where a rating changes and nobody can see that it did.

What moves a rating

A status changes when one of three things happens: a scheme that was hollow becomes genuinely sittable, an accreditation that was claimed gets confirmed or fails to check out against the accreditor’s own register, or someone tells us we have it wrong and turns out to be right. All three get logged the same way, dated, in the open.

What this is not

A rubric is not a guarantee. Our standings are informed opinion, built from public information and checked by hand, not an official determination, a legal judgment, or a certification of our own. We are not affiliated with, or endorsed by, any accreditation body or any credential we describe. For anything that affects a real decision, use the primary source link on the entry itself, not just our read of it.

Who funds this, and who never does

A rubric is only as trustworthy as the hand applying it, so here is exactly where our money comes from, and where it never does.